1APRIL 2021 T U R N E R E T A L . 2653

Extreme Temperatures in the Antarctic

a a a a a a JOHN TURNER, HUA LU, JOHN KING, GARETH J. MARSHALL, TONY PHILLIPS, DAN BANNISTER, AND a STEVE COLWELL a British Antarctic Survey, Natural Environment Research Council, Cambridge, United Kingdom

(Manuscript received 13 July 2020, in final form 6 November 2020)

ABSTRACT: We present the first Antarctic-wide analysis of extreme near-surface air temperatures based on data collected up to the end of 2019 as part of the synoptic meteorological observing programs. We consider temperatures at 17 stations on the Antarctic continent and nearby sub-Antarctic islands. We examine the frequency distributions of temper- atures and the highest and lowest individual temperatures observed. The variability and trends in the number of extreme temperatures were examined via the mean daily temperatures computed from the 0000, 0600, 1200, and 1800 UTC ob- servations, with the thresholds for extreme warm and cold days taken as the 5th and 95th percentiles. The five stations examined from the Antarctic Peninsula region all experienced a statistically significant increase (p , 0.01) in the number of extreme high temperatures in the late-twentieth-century part of their records, although the number of extremes decreased in subsequent years. For the period after 1979 we investigate the synoptic background to the extreme events using ECMWF interim reanalysis (ERA-Interim) fields. The majority of record high temperatures were recorded after the passage of air masses over high orography, with the air being warmed by the foehn effect. At some stations in coastal East Antarctica the highest temperatures were recorded after air with a high potential temperature descended from the Antarctic plateau, resulting in an air mass 58–78C warmer than the maritime air. Record low temperatures at the Antarctic Peninsula stations were observed during winters with positive sea ice anomalies over the Bellingshausen and Weddell Seas. KEYWORDS: Atmosphere; Antarctica; Antarctic Oscillation; Climate variability; Temperature

1. Introduction (63.48S, 57.08E; 13 m MSL) near the northern tip of the Antarctic Peninsula, recorded a temperature of 17.58C, which was the highest Antarctica is a land of extremes. It is the highest, coldest, accepted temperature measured on the Antarctic continent and driest, and windiest continent on Earth, but with a wide range nearby islands (Bozkurtetal.2018; Skansi et al. 2017). More re- of climatic conditions from the relatively warm, maritime cently, Esperanza reported a temperature of 18.38Con6February northern part of the Antarctic Peninsula to the frigid plateau of 2020, although this measurement has yet to be verified. East Antarctica (King and Turner 1997). These four extreme events have received extensive publicity, but Vostok station [78.58S, 106.98E; 3488 m above mean sea level there has been little research into extreme temperatures at other (m MSL)] on the East Antarctic plateau recorded the lowest Antarctic locations. A record high temperature at South Pole in surface temperature observed on Earth when, on 21 July 1983, the 1957 was examined by Alvarez and Lieske (1960) and attributed to temperature fell to 289.28C. Analysis of this event (Turner et al. warm-air advection from the coastal region over several days. A 2009) showed that the very low temperature occurred as a result of further event that gave record high temperatures at several stations the isolation of the station from relatively warm maritime air in 1978 was documented by Sinclair (1981), and again linked to masses for a period of 10 days. However, it was suggested that advection of warm air into the interior of the continent. More re- even lower temperatures could occur at higher elevations, such as cently, studies have shown that such intrusions of maritime air onto Dome A (4093 m MSL; 80.378S, 72.358E) and these have subse- the Antarctic plateau can be well represented by numerical analyses quently been identified using satellite observations and auto- (Nicolas and Bromwich 2011), indicating that the operational an- matic weather station (AWS) data (Scambos et al. 2018). alyses can be of value in investigating extremes. At the other extreme, it has recently been confirmed (Skansi Orcadas station on the South Orkney Islands northeast of et al. 2017) that the highest surface temperature measured south of the tip of the Antarctic Peninsula and north of the Weddell Sea 608Swas19.88C, recorded at Signy station (60.78S, 45.68W; 5 m has Antarctica’s longest meteorological record, starting in MSL), South Orkney Islands, on 30 January 1982. Analysis of this 1903. Although the early observations have yet to be digitized, event (King et al. 2017) showed that the high temperature occurred daily mean temperatures are available and trends in these when warm midtropospheric air descended to the station during a mean values and the extremes have been investigated by foehn event as a northerly airstream flowed over the high orog- Zazulie et al. (2010). They found that since 1950 there had raphy of Coronation Island. On 24 March 2015, Esperanza station been a reduction in the number of cold extremes, particularly in fall and winter. The positive trend in the summer warm

Denotes content that is immediately available upon publica- tion as open access. This article is licensed under a Creative Commons Attribution 4.0 license (http://creativecommons. Corresponding author: John Turner, [email protected] org/licenses/by/4.0/).

DOI: 10.1175/JCLI-D-20-0538.1

Ó 2021 American Meteorological Society Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2654 JOURNAL OF CLIMATE VOLUME 34 extremes since 1970 was twice the magnitude of that in the Pole and Vostok station. We have not used the measurements earlier part of the record, indicating the potential impact of made by AWSs since these systems are unattended for long stratospheric ozone depletion. periods and there are questions regarding the accuracy of the A knowledge of extreme conditions, including their fre- temperature observations in extreme conditions. quency and the synoptic background to the events, is important Quality control of the observations is extremely important in for ensuring the physical maintenance and safety of the sta- assessing extreme events, so this topic is discussed in section 2, tions, but is also of great research interest, as extreme condi- along with the means by which we determined the thresholds tions can contribute significantly to the breakup of ice shelves for extreme conditions. Section 3 examines the frequency during the summer months (Scambos et al. 2004). High tem- distributions of temperatures from the stations and the factors perature events can be associated with significant amounts of that influence the shape of the distributions. The highest and precipitation, which are important in the interpretation of the lowest temperatures recorded at the stations are discussed in signals in ice cores (Schlosser et al. 2016; Turner et al. 2019a). section 4, along with the synoptic environments in which these Extreme atmospheric conditions can also have an impact on occurred. The variability and trends in the occurrence of ex- the formation and melting of sea ice, which in turn can affect treme warm and cold days over the full length of the records the ocean environment (Turner et al. 2017). are considered in section 5. Conclusions are presented in There are different interpretations as to what constitutes an section 6. extreme event. Determining a suitable definition often depends on the application or reason for considering meteorological pa- 2. Data and methods rameters that are well beyond the mean. For applications con- cerned with economic and social impacts, a temperature threshold We use the synoptic observations of surface temperature made as that leads to physical hardship to certain sections of the commu- part of the routine meteorological observing programs carried out at nity may be appropriate. In tropical areas a threshold temperature the stations up to the end of 2019. In the early parts of the records the of 358 or 408C can be appropriate because of the societal impacts measurements were made manually by observers at the main syn- (Zhai and Pan 2003), with such temperatures occurring on several optic hours, with the data being collected every 3 or 6 h, or more days or more being very important. In the Antarctic, tempera- infrequently depending on the availability of staff. Gradually, there tures above the freezing point are particularly important because was a transition to the automated collection of data, which allowed of the resultant ice melt and impact on the terrestrial biota (Wake more frequent observations to be obtained, with hourly or more and Marshall 2015). Other investigations of extremes have fo- frequent data often being archived. The observations collected were cused on temperatures above or below a specified percentile, with brought together as part of the Scientific Committee on Antarctic the 5th and 95th percentiles often being used (Cardil et al. 2014). Research (SCAR) Reference Antarctic Data for Environmental In other cases an extreme temperature has been taken as a value Research (READER) project (Turner et al. 2004), with the data beyond 2 standard deviations of the mean (Vavrus et al. 2006), but being obtained from the operators of the national Antarctic pro- this is of most value if the data have a normal frequency distri- grams. The data were subject to a rigorous quality control procedure bution, which, as we show in this paper, is not the case for most (see Turner et al. 2004) and the computed monthly mean values up Antarctic temperature data. to 2015 form the basis of the online SCAR READER database While there have been a number of climatological studies into (https://legacy.bas.ac.uk/met/READER/).Since2015wehaveused extreme temperatures in the extrapolar regions, very few have the reports from the station operators where possible but if these considered Antarctica. Wei et al. (2019) examined changes in ex- were not available, we used the synoptic reports distributed over the treme temperatures over 1970–2013 using daily maximum and Global Telecommunications System. For these observations, we minimum temperatures, but these quantities are not generally applied the same quality control procedure as applied to the earlier available digitally for Antarctic stations prior to 2000: thus, data data. Briefly, this involved flagging erroneous observations by from only four stations were available to calculate trends for 1970– comparing the measurements with the climatological mean for the 2000. In addition, they did not examine the synoptic background to location and identifying outliers and unrealistic rapid changes in extreme temperature events. Here we therefore consider Antarctic temperature. All record low and high temperatures were examined extreme temperatures using all the synoptic observations available in detail by following the evolution of temperature and other me- from 17 stations with long records. Clearly, slightly higher or lower teorological parameters through the event and, for post-1979 rec- temperatures will have occurred between the standard observing ords, examining the synoptic analyses from the ECMWF interim times and extremes outside the normal reporting times are noted reanalysis (ERA-Interim). Details of the background to specific where known. However, the use of the temperatures from the extreme temperature events are provided in the text. standard reporting hours allows the investigation of the frequency We have investigated extreme temperatures at 17 stations of extreme temperatures and the synoptic environment at the (the locations are listed in Table 1 and indicated in Fig. 1) that times of record temperatures because of the availability of routine have long, year-round records and that are distributed fairly analyses. In addition, it allows the investigation of whether the evenly around the Antarctic coastline. Many of the records number of extreme warm and cold days has changed since the start around the time of the International Geophysical Year stations were established. (IGY) in the late 1950s, with the shortest record examined Most of the stations we consider are located in the Antarctic being from Neumayer station, which starts in January 1981. We coastal region; however, there are two long records from the have not merged any records from nearby stations, but there interior plateau from Amundsen–Scott station at the South have inevitably been small changes in the locations of some

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2655

TABLE 1. Stations considered in this study. Note that meteorological observations were first collected at Orcadas in 1904 but have only been digitized for the period since January 1956.

Station Latitude Longitude Elevation (m) Data period Orcadas 60.738S 44.738W 6 Jan 1956–Dec 2019 Bellingshausen 62.188S 58.958W 16 Mar 1968–Dec 2019 Esperanza 63.408S 57.008W 13 Jan 1957–Dec 2019 Marambio 64.248S 56.638W 198 Jan 1971–Dec 2019 Vernadsky 65.258S 64.258W 11 Jan 1947–Dec 2019 Casey 66.288S 110.538E 41 Feb 1969–Dec 2019 Mirny 66.568S 93.008E 30 Feb 1956–Dec 2019 Dumont d’Urville 66.678S 140.008E 43 Apr 1956–Dec 2019 Rothera 67.578S 68.138W 16 Mar 1976–Dec 2019 Mawson 67.608S 62.878E 16 Feb 1954–Dec 2019 Molodezhnaya 67.678S 45.838E 40 Mar 1963–Jun 1999, Jan 2010–Dec 2019 Davis 68.588S 77.978E 13 Feb 1957–Dec 2019 Syowa 69.008S 39.588E 21 Feb 1957–Dec 2019 Neumayer 70.658S 8.48W 50 Jan 1981–Dec 2019 Novolazarevskaya 70.778S 11.818E 99 Feb 1961–Dec 2019 Vostok 78.468S 106.848E 3488 Jan 1958–Dec 2019 Amundsen–Scott 90.008S — 2800 Jan 1957–Dec 2019 stations because of rebuilds and relocations (where available, situation in which extreme temperatures occurred for the period details are provided in the metadata section of the READER from 1979 as the reanalysis fields are well constrained by satellite website at https://legacy.bas.ac.uk/met/READER/metadata/ data from this date onward. For this we used the fields from the metadata.html). However, these should not have had a major ERA-Interim (Dee et al. 2011), which have a grid spacing of impact on temperatures at most stations in the way that some ;70 km. A number of studies have conducted intercomparisons wind records have been affected by relocations. We have not used of the various reanalysis datasets and concluded that the ERA- data from Halley station on the Brunt Ice Shelf on the eastern side Interim data are the best for depicting recent Antarctic climate of the Weddell Sea in this study as the station has been moved (Bracegirdle and Marshall 2012). inland five times during its lifetime in order to maintain a safe There are many earlier studies that have examined changes distance from the ice edge. This change in distance from the rel- in the daily maximum and minimum temperature either atively warm ocean has affected the homogeneity of the tem- globally or for a particular region (e.g., European Academies perature series, with at least one large jump in the record and Science Advisory Council 2013). However, as noted by Wei other smaller changes. The station moves will also have affected et al. (2019), such data are not readily available in digital form the record of extreme temperatures at the station. prior to 2000 for many of the Antarctic stations and where data Making instrumental measurements at any location presents have been digitized they do not cover the full length of the records. challenges, especially during extreme conditions, and great care We have therefore elected to examine variability and change in the needs to be exercised regarding the instruments used, the daily mean temperature, which we have calculated from the 0000, shielding of sensors, and the selection of a location for the in- 0600, 1200, and 1800 UTC observations. The daily mean temper- strument screen. In the Antarctic, most operators use a standard ature has been used in many glaciological investigations where meteorological screen, which is located at a suitable site close to there is a requirement for data on positive degree-days (Barrand the main station. Such instruments provide satisfactory data under et al. 2013). Here we only computed a daily mean temperature if all most conditions. However, it has been shown that in the Antarctic four of the main 6-hourly synoptic observations were available. To summer during periods of low wind speed and clear skies, tem- identify the thresholds for extreme cold and warm days we have peratures measured in a standard screen can have a positive bias used the 5th and 95th percentiles of the temperature distribution, a (Burton 2014; Genthon et al. 2011). Scatter diagrams of high method that has been used in many earlier investigations of ex- temperatures against wind speed do show that a number of the treme temperatures [see references in IPCC (2012)]. As the tem- record high temperatures reported in the station data occurred perature records from the stations are of different length, we have when the wind speed was very low. We have therefore provided determined the 5th and 95th percentiles for each station using two record high temperatures for each station—one being the temperatures for the 30-yr period 1980–2009 and then applied highest temperature recorded regardless of wind speed, and the them over the full length of the records. Since the number of days second for the highest temperature measured when the wind in each year with valid data is variable, we show the percentage of 2 speed was 5 knots (1 kt ’ 0.51 m s 1)orgreater. days with extreme high and low temperatures rather than the ac- Although some synoptic charts for high southern latitudes tual number of days. We only compute the percentage of days with were drawn manually around the time of the IGY, there were extreme temperatures if there are at least 90% of daily mean very few observations over the ocean, and it is felt that the fields temperatures available in a year. are not of high value in the analyses of the broadscale synoptic We calculated the trend in the percentage of days each year environment. We have therefore only considered the synoptic that had extreme temperatures using ordinary least squares

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2656 JOURNAL OF CLIMATE VOLUME 34

FIG. 1. Histograms of all 6-hourly temperature observations for the period 1980–2009 for the extended summer half year (November– April; light red) and the extended winter half year (May–October; blue). The overlap between the summer and winter histograms is indicated by dark red. The distribution of temperatures for the whole year are indicated by the black line. The figure also indicates the locations of the stations considered in this study with the red dots identifying the six stations considered in detail. regression techniques, while using the nonparametric Mann– icd/gjma/sam.html) and the Climate Prediction Centre Niño-3.4 Kendall test to calculate trend significance (Alexander and seasonal temperature anomalies (http://www.cpc.ncep.noaa.gov/ Arblaster 2009). This test is widely used for evaluating the data/indices/). The percentage of warm and cold extremes each presence of monotonic trends and makes no assumption for year has respectively been correlated with the summer and winter normality, only that the data are temporally independent, so is SAM index and Niño-3.4 temperature anomaly. appropriate for the extreme temperature data. To investigate the relationships between the percentage of 3. The temperature distributions at the stations warm and cold extremes, and the major modes of climate vari- ability we have used the index of the southern annular mode We first examine the frequency distributions of tempera- (SAM) developed by Marshall (2003) (http://www.nerc-bas.ac.uk/ tures via histograms of all the 6-hourly synoptic observations

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2657 from the stations over the period 1980–2009 (Fig. 1). Data are of year. Sea ice is particularly variable in the middle of winter presented for the extended summer half year (November– when a polynya-like area of open water is found close to the April), the extended winter half year (May–October), and the station in some years (Turner et al. 2013). With open water near year as a whole. We will focus particularly on data from six Vernadsky and air masses arriving from the north, midwinter stations covering the broad climatic regions across Antarctica. temperatures can be several degrees above freezing (Turner These are Orcadas at a maritime Antarctic island location, et al. 2019b). Marambio and Vernadsky (formerly Faraday) on the eastern There are large differences in the shape of the annual temperature and western sides of the Antarctic Peninsula, respectively, distribution for stations around the coast of East Antarctica. These Neumayer on the coast of the eastern Weddell Sea, Mawson on vary from a single peak (Syowa) to a broad, flat distribution of the coast of East Antarctica, and Amundsen–Scott on the high temperature in the middle of the range (Molodezhnaya) to a clear interior plateau. bimodal structure (Dumont d’Urville). The distributions for the ex- All the non-plateau stations have a negatively skewed an- tended winter 6 months (Fig. 1, blue distributions) are quite similar nual temperature distribution with a long cold tail. As dis- for the majority of the stations, with a single peak in the distribution, cussed below, this is because the extreme high temperatures although the kurtosis (sharpness of the peak) varies. The exception is are associated with transient warm-advection events (e.g., Casey, which has a more bimodal winter temperature distribution, Bozkurt et al. 2018; King et al. 2017). In contrast, the very cold with the main peak at 2108C and a secondary peak close to 2188C. temperatures are associated with periods of isolation from This form of distribution is most pronounced in May and September warmer air masses, when there is often extensive sea ice cover, when the circumpolar trough is farther south and there is a minimum and when the temperatures can drop to very low values over in the mean sea level pressure (MSLP) field off the station (van den extended periods (Ding et al. 2020; Turner et al. 2013; Turner Broeke 1998).Thepeakinthedistributionat2108C is associated et al. 2009). The extended summer distributions of the coastal with the many cyclones that are found just off the coast in the area East Antarctic stations have a more negatively skewed distri- (Jones and Simmonds 1993), which bring relatively warm air from the bution than that for the extended winter. This is likely a result northtothestation.Thepeakat2188C is indicative of occasions of the extended summer period including both the months when there are few depressions off the coast and low wind speeds when there are periods of 24 h of sunlight, and times when that preserve the surface temperature inversion and lead to lower there is a period of darkness with lower temperatures. temperatures—the mean 10-m wind speed for occasions when sur- 2 For stations with a strong maritime influence, the annual face temperature is in the range 217.08 to 219.08Cis4.1ms 1, 2 temperature distributions have a single sharp peak close to compared to 9.6 m s 1 for temperatures of 29.08 to 211.08C. The freezing point as a result of the cold ocean temperatures that differences in the winter temperature distributions are strongly are found throughout much of the year. This is the case with influenced by the different wind regimes at the stations associated Orcadas (Zazulie et al. 2010), which is the most northerly with local and broadscale factors. station examined (Fig. 1). The temperature distributions for the extended summer Vernadsky and Bellingshausen on the western side of the 6 months are mostly characterized by a single peak 18C below Antarctic Peninsula both have similar temperature distribu- freezing point because of the extensive areas of open water just tions to Orcadas, reflecting the frequency of ice-free conditions north of the stations during this half year. But as with the dis- at these locations (Turner et al. 2013). In the extended summer, tributions for the extended winter, the kurtosis varies. Syowa there is little sea ice near Vernadsky (Parkinson 2014) and the has a very narrow peak and a kurtosis of 4.13. In contrast, the strong maritime influence is reflected in the high proportion of distribution for Mawson has a broad tail of low temperatures temperatures close to the freezing point (Fig. 1). However, (the kurtosis is 2.78) because of the high frequency of cold, Rothera, in the more southerly part of the western Antarctic katabatic winds, even during the summer half year. Peninsula, has a slightly broader annual temperature distri- On the Antarctic plateau the two stations with long records both bution because of the greater frequency of sea ice close to the have annual temperature distributions with a marked bimodal station (Comiso and Nishio 2008). form, with the cold peak being more pronounced than the warm. A bimodal structure in the temperature distribution is apparent The mean annual cycle of temperature at both stations has a clear in the annual data from a number of stations, especially at more signal of the coreless winter with an extended winter of low tem- southerly latitudes. At Marambio (Fig. 1)itarisesbecauseofthe peratures and a short, relatively warm summer (Thompson et al. ice-free conditions over the extended summer giving a sharp peak, 1970). Because of the length of the coreless winter (approximately coupled with a broader distribution in winter when there is more March to October), the 6-month extended summer period we extensive and variable sea ice cover (Zwally et al. 2002). The two examine here (November to April) includes months of low tem- winter peaks correspond broadly to the frequent cold barrier peratures, so that the extended summer temperature distributions winds from the south (King and Turner 1997)andwesterlies for the two station have a bimodal distribution, which is slightly crossing the Antarctic Peninsula, many of which are warmed by more pronounced at Vostok. the foehn effect (Elvidge et al. 2016). A similar annual tempera- ture distribution is found at Esperanza, although the winter cold 4. The extreme temperatures peak is slightly less pronounced. The extended winter distribution of temperatures at Vernadsky We consider the record high and low temperatures recorded is broader than in the extended summer, reflecting the highly in the station synoptic observations (Table 2) and examine the variable nature of sea ice and atmospheric circulation at this time conditions that led to such extreme conditions. Composite

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2658 JOURNAL OF CLIMATE VOLUME 34

TABLE 2. The record high and low temperatures recorded at the stations up to the end of 2019.

Maximum temperature and the time Maximum temperature and the time and date with the wind speed 5 kt or Minimum temperature and the time Station and date stronger and date Orcadas 13.58C (1800 UTC 23 Dec 1987) 13.58C (1800 UTC 23 Dec 1987) 238.98C (0000 UTC 28 Jun 1972) Bellingshausen 11.28C (1200 UTC 29 Jan 1982) 11.28C (1200 UTC 29 Jan 1982) 228.58C (0600 UTC 5 Aug 1991) Esperanza 17.58C (0900 UTC 24 Mar 2015)a 17.58C (0900 UTC 24 Mar 2015) 231.88C (1200 and 1800 UTC 8 Jul 1975) Marambio 17.18C (2100 UTC 23 Mar 2015) 15.28C (1800 UTC 4 Jan 1979)b 236.88C (2100 UTC 25 Jul 1994) Vernadsky 10.98C (0000 UTC 25 Jan 1985 and 10.98C (0000 UTC 25 Jan 1985) 242.48C (0300 UTC 10 Aug 1958) 0100 UTC 31 Dec 1988) Casey 8.48C (0900 UTC 30 Jan 1991) 8.08C (0000 UTC 12 Jan 1980) 238.08C (2100 UTC 17 Jul 1976 and 0900 UTC 18 Jul 1986) Mirny 6.78C (0600 UTC 2 Jan 1998) 6.08C (0000 UTC 9 Jan 1960) 239.58C (0000 UTC 20 Aug 1956 and 0600 UTC 15 Aug 1982) Dumont d’Urville 8.78C (0600 UTC 2 Jan 2002) 8.68C (1200 UTC 30 Dec 2001) 235.28C (0000 UTC 18 Jul 1976) Rothera 8.78C (2000 UTC 20 Jan 2003) 8.08C (0600 UTC 8 Feb 1999) 239.58C (1200 UTC 12 Aug 1980) Mawson 9.08C (0900 UTC 9 Jan 1974) 9.08C (0900 UTC 9 Jan 1974) 235.38C (0300 UTC 16 Jul 1985) Molodezhnaya 7.88C (1200 UTC 9 Dec 1967) 7.88C (1200 UTC 9 Dec 1967) 239.28C (0600 UTC 31 Aug 1993) Davis 138C (0600 UTC 7 Jan 1974) 138C (0600 UTC 7 Jan 1974) 240.78C (0000 UTC 27 Apr 1998) Syowa 9.38C (0900 UTC 15 Jan 1969) 8.18C (1500 UTC 1 Dec 1997) 244.18C (0200 UTC 4 Sep 1982) Neumayer 4.48C (0600 UTC 28 Mar 1991) 4.48C (0600 UTC 28 Mar 1991) 249.88C (0000 UTC 8 Jul 2010) Novolazarevskaya 9.78C (1200 UTC 3 Feb 1996) 9.78C (1200 UTC 3 Feb 1996) 238.88C (1800 UTC 8 Aug 1965) Vostok 214.48C (0600 UTC 6 Jan 1974) 214.48C (0600 UTC 6 Jan 1974) 289.28C (0300 UTC 21 Jul 1983) Amundsen–Scott 212.38C (1000 UTC 25 Dec 2011) 212.38C (1000 UTC 25 Dec 2011) 281.78C (0000 UTC 23 Jun 1982) a A measurement of 18.38C on 6 February 2020 has yet to be verified. b Note that a temperature of 15.68C was reported at 1800 UTC 6 February 2020. anomalies in the MSLP (upper-air geopotential height fields In the following we examine the conditions that led to ex- for Amundsen–Scott) are presented for the days when the treme temperatures at the six stations representative of dif- temperatures were beyond the thresholds listed in Table 3 ferent climatological zones within the Antarctic. during the post-1979 period. a. Orcadas Extreme high temperatures clearly occur during the extended austral summer, with most having been recorded in January, Orcadas, in the maritime Antarctic South Orkney Islands, although the highest temperatures at Neumayer, Marambio, and has a climate that is strongly influenced by the many eastward Esperanza were all observed in March. High temperatures not moving storms over the Southern Ocean, along with cold air linked to foehn dynamics are unlikely before December because masses arriving from the cold, sea ice-covered Weddell Sea. The of the extensive sea ice that influences temperatures at many highest temperatures at the station have been associated with stations. While there is a general north–south decrease in the strong westerly flow between an area of high pressure over the record high temperatures that have been observed, Orcadas has South Atlantic and low pressure in the Weddell Sea (Fig. 2a). In not recorded a temperature as high as Esperanza and Marambio this situation, the foehn effect increases the temperature as the since air masses arriving at the station are not warmed to such an air passes over the high orography of Coronation Island to the extent by the foehn effect as occurs at the stations in the west of the station [e.g., see King et al. (2017) for a detailed northeastern part of the Antarctic Peninsula. The exact location investigation of a foehn wind affecting conditions on nearby of a station in relation to the local orography is extremely im- Signy Island]. Low MSLP in the Bellingshausen Sea is associated portant in determining whether it experiences foehn winds, as with La Niña conditions in the tropical Pacific Ocean so the evidenced by the much higher temperatures that have been number of warm events at Orcadas has an anticorrelation recorded at Signy compared to the nearby Orcadas station. of 20.3 (p , 0.05) with the Niño-3.4 index. The highest tem- Extreme low temperatures are invariably recorded during the perature recorded was 13.58C at 1800 UTC 23 December 1987, extended winter season, with most occurring in July and August, with the temperature increasing from around 08C to this record although some were recorded in June and September. Conditions level in 12 h before decreasing to 28C 12 h later. This short, rapid conducive to very low temperatures are isolation from relatively incursion of warm air, which was not observed at the nearby mild midlatitude air masses for extended periods and extensive Signy station (Signy is not directly in the lee of Coronation sea ice cover in the vicinity of the station to limit the flux of heat Island during westerly winds), is characteristic of other very high and moisture from the ocean. Such conditions are found along the temperature events occurring at the station. eastern side of the Weddell Sea, so that Neumayer has recorded The presence or absence of sea ice is particularly important extreme low temperatures below anything observed at the sta- in dictating the temperatures at the more northerly stations, as tions around the coast of East Antarctica. extensive sea ice limits the flux of heat from the ocean. At

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2659

TABLE 3. The upper and lower thresholds for extreme daily mean temperatures for the 17 stations based on the 5th and 95th percentiles determined from the daily mean temperature (mean of the temperatures at the four main reporting hours) for 1980–2009. We also include the range of months over which the extremes occurred in the period 1980–2009.

Station Upper threshold Range of months for upper extremes Lower threshold Range of months for lower extremes Orcadas 2.58C January–December 217.38C May–October Bellingshausen 2.78C November–April 211.38C April–October Esperanza 3.78C January–December 218.08C April–October Marambio 2.28C January–December 223.48C April–October Vernadsky 2.38C August–June 212.88C May–November Casey 0.68C November–March, June, August 221.98C March–October Mirny 20.98C November–March 223.38C April–October Dumont d’Urville 20.18C November–March 222.18C April–October Rothera 2.38C October–June 217.38C April–October Mawson 0.78C November–February 224.08C April–October Molodezhnaya 20.18C November–May 223.58C April–October Davis 1.58C November–April 224.28C April–October Syowa 0.18C November–March, May 224.48C April–October Neumayer 22.78C November–April 232.98C April–October Novolazarevskaya 0.78C November–February 222.38C April–October Vostok 230.08C November–February 274.78C April–September Amundsen–Scott 225.98C November–February 268.38C March–October

Orcadas the ice extent around the islands is highly variable in winter. Esperanza its highest verified temperature of 17.58C, which has This is reflected in the midwinter temperature distribution (not been described in detail by Skansi et al. (2017). shown), which has peaks just below freezing and around 2158C, It is surprising that temperatures as high as 10.38C (at corresponding to ice-free and ice-covered conditions, respectively. 0000 UTC 15 July 2010) have been recorded at Marambio in The lowest temperatures were associated with positive sea ice extent midwinter; however, Kirchgaessner et al. (2019) showed that anomalies, with the ice extending northward from the Weddell Sea the temperatures on the eastern side of the Antarctic Peninsula to the latitude of the South Orkney Islands. The mean MSLP during foehn events show little variability over the year. The anomalies for the occasions with the lowest temperatures at synoptic environment when such high winter temperatures oc- Orcadas (Fig. 3a) has a negative pressure anomaly to the east of cur consists of a warm ridge extending southward from southern the islands and positive anomaly over the Antarctic Peninsula. South America toward the Antarctic Peninsula, which brings This pattern favors a strong southerly flow of cold air out of the warm South Pacific air over the eastern Bellingshausen Sea. The Weddell Sea, in addition to advecting sea ice toward the region. air then crosses the Antarctic Peninsula as a strong northwest- erly with temperatures increased by the foehn effect. b. The eastern Antarctic Peninsula Generally, the lowest temperatures on the eastern side of the Temperatures on the eastern side of the Antarctic Peninsula Antarctic Peninsula are associated with a south to southeasterly are strongly influenced by the high orography to the west, with airflow from the cold Weddell Sea when there is low pressure over the extreme high temperatures associated with strong westerly the northwestern Weddell Sea (Fig. 3b). Under these conditions, winds. The mean MSLP anomaly for the highest 5% of tem- there is greater sea ice advection than normal toward the north- peratures at Marambio (Fig. 2b) is negative in the region of the west Weddell Sea, limiting the flux of heat from the ocean. The Amundsen Sea low, with a positive anomaly over the south- number of cold events is significantly anticorrelated (20.34, p , western Atlantic/Falkland Islands, implying that the belt of 0.05) with the SAM index as weaker westerly winds are associated strong westerlies is displaced toward the south. This results in with more meridional flow out of the Weddell Sea. The lowest relatively warm air masses being further warmed by the foehn temperature recorded at Marambio (236.88C, at 2100 UTC effect as they pass over the peninsula before arriving at the 25 July 1994) occurred toward the end of 6 days of decreasing stations (Cape et al. 2015; Elvidge et al. 2016; Marshall et al. temperature as high pressure built across the region from an an- 2006). The highest temperature observed at Marambio was ticyclone over the Bellingshausen Sea at the same time as a very 17.18C at 2100 UTC 23 March 2015, during a period of strong deep low formed over the northeast Weddell Sea. In the marked westerly winds involving warm air from the South Pacific. MSLP gradient across the Weddell Sea very cold air was drawn However, at the time of the record temperature, the wind toward the northwest Weddell Sea with the ERA-Interim re- speed had dropped to only 1 kt, so the temperature may have analysis indicating that surface temperatures of ,2358C occurred been higher because of enhancement through strong solar ra- across a large part of the Weddell Sea. diation onto the instrument enclosure. The highest tempera- c. The western Antarctic Peninsula ture recorded at the station with a wind speed of more than 5 kt was 15.28C at 1800 UTC 4 January 1979 during another strong The Antarctic Peninsula extends northward from the main westerly wind event involving air from the South Pacific. The body of the Antarctic and the temperatures at the stations are period of strong westerlies over 23–24 March 2015 also gave strongly influenced by synoptic-scale weather systems within

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2660 JOURNAL OF CLIMATE VOLUME 34

FIG. 2. The composite MSLP (500-hPa geopotential height for Amundsen–Scott) anomalies for the days during 1979–2019 when the temperatures at the stations were above the high temperature thresholds listed in Table 3, shown for (a) Orcadas, (b) Marambio, (c) Vernadsky, (d) Neumayer, (e) Mawson, and (f) Amundsen–Scott. The base period for the anomalies was the 31-day centered running mean (excluding the central day itself). The stippling indicates where the anomalies are statistically significant at p , 0.05. The location of each station is shown by a white dot.

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2661

FIG.3.AsinFig. 2, but for occasions when the temperatures were below the cold temperature threshold.

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2662 JOURNAL OF CLIMATE VOLUME 34 the climatological circumpolar trough that extends across 608–708S. d. The eastern Weddell Sea The highest temperatures recorded at Vernadsky have been asso- Neumayer experiences a climatological easterly wind as it is ciated with large negative MSLP anomalies over the Bellingshausen located on the southern side of the Circumpolar Trough. The Sea (Fig. 2c), with the lows in this area advecting warm air masses highest temperatures recorded here generally occurred with down the peninsula. Several factors are important in determining the lower pressure than normal over the northeastern Weddell Sea temperatures at Vernadsky, including the location of the low in the and high pressure in the Circumpolar Trough to the east Bellingshausen Sea, the pressure of the high in the Weddell Sea, and Greenwich Meridian. This synoptic situation favors the ad- the origin of the air mass drawn toward the western side of the vection of warm air masses toward the Antarctic coast down peninsula. In contrast, the lowest temperatures occurred with posi- the Greenwich meridian (Fig. 2d). When Neumayer recorded tive MSLP anomalies over the Bellingshausen Sea and negative its highest temperature of 4.48C at 0600 UTC 28 March 1991 anomalies in the Weddell Sea, giving cold-air advection over the there was a complex area of low pressure to the northwest of region (Fig. 3c). The MSLP anomalies for high temperatures the station that drew warm air from the north, but with only a resemble a La Niña–like pattern (the anticorrelation between the ridge of high pressure to the east of the Greenwich meridian. number of warm events and the Niño-3.4 index is 20.34, p , 0.01), The lowest temperatures at Neumayer were recorded with a while the mean MSLP anomalies for low temperatures resemble an negative pressure anomaly over Dronning Maud Land and El Niño–like pattern. However, the warm extremes do not prefer- high MSLP in the Weddell Sea (Fig. 3d). This synoptic situa- entially occur during La Niña conditions, nor the cold extremes tion results in cold air masses arriving at the station from a during El Niño events. This highlights the different impacts of short- generally southerly direction. The lowest temperature ob- term, high-amplitude weather events and the lower-frequency served at the station was 249.88C at 0000 UTC 8 July 2010. The preferential climate-mode type patterns that occur when averaged long-term mean July temperature at the station is 224.88C, and over El NiñoandLaNiñaevents. 2010 had the third coldest July in the record starting in 1982. The highest temperature in the Vernadsky record was The record low temperature occurred in the middle of a cold 10.98C, which was measured at 0000 UTC 25 January 1985 and spell over the first 10 days of the month characterized by a 0100 UTC 31 December 1988. The 1985 record temperature trough of low pressure extending down Coats Land that ad- occurred when a very deep depression with a central MSLP vected cold air toward Neumayer from the south. The record of ,960 hPa was located over the southern Bellingshausen Sea low temperature on 8 July occurred as the wind speed dropped giving strong warm-air advection down the western side of the to 7 kt, enhancing the strength of the surface inversion. peninsula. This was a relatively short-lived event, with tem- peratures increasing from 28C to the record high and back to e. East Antarctica less than 48C within 48 h. High temperatures at Bellingshausen and Rothera are also The composite MSLP anomaly field for the highest 5% of often associated with deep depressions to the west of the penin- temperatures at Mawson has negative values close to 908Eand sula that advect warm air from the north. This scenario gave the positive values near 308E, suggesting that the highest tempera- record high Bellingshausen temperature of 11.28C at 1200 UTC tures at the station are associated with broadscale southerly flow 29 January 1982, although the advection of warm air was en- off the continent (Fig. 2e). The warm events also occur with hanced by a ridge of high pressure in the South Atlantic. However, positive MSLP–upper-level height anomalies over the interior of the record high temperature at Rothera of 8.78C at 2000 UTC the Antarctic, which is consistent with the significant anti- 20 January 2003 occurred with high MSLP over the area, although correlation (20.53, p , 0.01) between the number of warm the low wind speed of 4 kt coupled with cloud-free conditions may events each year and the SAM index. The positive MSLP have resulted in the thermometer reading too high. In contrast, anomaly inland of the station (also apparent in the surface the temperature of 8.08C at 0600 UTC 8 February 1999 occurred pressure) will increase the strength of the coastal easterly wind, with a strong easterly flow across the peninsula, suggesting and also feed air from the interior down into the coastal zone. warming of the air as a result of the foehn effect. The plateau air has a high potential temperature and as it de- Record low temperatures at stations on the western side of scends as a low-level flow will warm at the dry adiabatic lapse the Antarctic Peninsula are associated with strong southerly rate (DALR) and introduce air that is 58–78C warmer than that flow, bringing cold air masses across the region, often because normally found in the coastal region. The warm events are also of a couplet of relatively high pressure to the west of the pen- associated with a sharp trough to the east of Mawson, suggesting insula and low pressure to the east (Fig. 3c). The Rothera record that air is advected up on to the plateau before its final descent to low temperature (1200 UTC 12 August 1980, 239.58C) occurred the station. If this air is of maritime origin its forced ascent to with a very deep low pressure system in the Weddell Sea with a higher elevations inland of the coast will result in cooling at the central pressure of ,940 hPa, which contributed to the strong saturated adiabatic lapse rate. Subsequent descent of dry air to pressure gradient across the Antarctic Peninsula. The winter of the coast will warm the air mass at the DALR, giving a net 1980 was characterized by a large positive sea ice anomaly on the warming before the air arrives at the station. The process is western side of the peninsula that was established as a result of similar to that of the foehn effect, which is so important in giving the greater frequency of southerly flow. The record low tem- high temperatures in the Antarctic Peninsula. perature at Bellingshausen (0600 UTC 5 August 1991, 228.58C) The highest temperature recorded at Mawson during the also occurred during a winter of positive sea ice anomaly over period for which we have reliable atmospheric analyses was the Bellingshausen and Weddell Seas. 8.18C at 1200 UTC 5 December 1989. The MSLP anomaly

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2663 during this event was very similar to that shown in Fig. 2e, station was 212.38C at 1000 UTC 25 December 2011, with the with a deep low near 908E and a sharp trough just to the east of temperature being recorded during an intrusion of very warm Mawson. Back trajectories from Mawson indicate that the air air from the Bellingshausen Sea sector that lasted several days. responsible for the record high temperature originated on the Very low temperatures at Amundsen–Scott are a result of Antarctic plateau and descended to the coastal region near isolation from relatively warm air masses originating in lower 1008E, before being carried toward the station in the strong latitudes, particularly when the SAM is positive (the correlation of coastal easterlies. In addition, satellite imagery shows exten- the number of warm events with the SAM index is 0.37, p , 0.01). sive cloud over the Amery Ice Shelf ahead of the relatively The mean 500-hPa geopotential height anomaly for the coldest warm plateau air, suggesting an involvement of both maritime days (Fig. 3f) has small-amplitude planetary waves and a center of air aloft associated with the coastal depression and the near- negative anomaly located close to the station. This is broadly the surface air descending from the plateau. same scenario as occurred during the record low temperature While Fig. 2e shows the mean MSLP anomaly field for the recorded at Vostok station in 1989 (Turner et al. 2009). highest 5% of temperatures at Mawson, examination of individual The lowest temperature at Amundsen–Scott station was 281.78C cases of warm events at the station shows some different scenarios recorded at 0000 UTC 23 June 1982. During the period leading up associated with high temperatures. For example, the temperature of to this event there was little thermal advection into the interior of 7.38C recorded at 0900 UTC 19 December 1982 occurred when the continent and the circulation at 500 hPa consisted of a circular there was a marked ridge to the west of the station that advected air flow that kept warm air masses from the interior of the Antarctic. inland close to 458E, before it descended to Mawson along 608E. However, this regime broke down within two days, with the Flow down from the interior is also a feature of high temperatures temperature increasing rapidly to 2538C. at Dumont d’Urville station, with forced ascent of air up onto the plateau and then descent down the glacial valleys. However, the 5. Variability and change in the frequency of extreme highest temperatures recorded at some other stations around the daily mean temperatures coast of East Antarctica, such as Davis and Casey, are associated with warm air advected around a low north of the station and We have examined variability and change of extreme high carried directly to the station in the strong easterly flow. and low temperatures via the number of daily mean tempera- Low temperatures at Mawson are associated with relatively tures that exceeded the 5th and 95th percentiles. The warm and high MSLP to the north of the station and low pressure over the cold threshold values (Table 3) were computed using all the Antarctic interior (the correlation of the number of cold events daily mean values over 1980–2009. To take account of the with the SAM is 0.38, p , 0.01), a situation that will weaken the variability in the number of valid daily mean temperatures coastal easterlies, with the airflow around the low drawing cold between years, we consider the percentage of days each year air from the interior (Fig. 3e). The relatively high pressure will with extreme warm or cold days. promote generally cloud-free conditions and give greater The time series of the percentage of days that were classified as cooling through emission of longwave radiation. However, the warm or cold for each year over the full length of the records for speed of the southerly wind must not be too strong or else the the six stations considered in detail are presented in Figs. 4 and 5. surface temperature inversion will break down. We only consider years when there were daily mean temperatures The lowest temperature recorded at the station was 235.38C available for 90% of the days. In the following we examine the at 0300 UTC 16 July 1985. The synoptic situation when this variability and change in the percentage of warm and cold ex- record low temperature was recorded was rather different to treme days for stations in the six regions discussed earlier. that for many of the other very low temperatures. On this a. Orcadas occasion a deep depression with a central pressure of ,956 hPa was present in the Antarctic coastal region close to 1208E and a High temperatures at Orcadas occur with strong westerly winds, marked trough extended to almost 808S over the Amery Ice with the foehn effect increasing temperatures as the air passes over Shelf. This resulted in a southerly flow of cold air from the high orography. Over recent decades the strength of the westerlies interior to the coastal region, although at the time of the lowest has increased in summer, at least in part because of the ozone hole 2 temperature the wind speed had dropped to just 1 m s 1. (Abram et al. 2014). This has shifted the SAM to being predomi- nantly in its positive polarity (Marshall 2003). The positive trend in f. The Antarctic Plateau 2 the percentage of warm days at Orcadas (0.60% decade 1, p , High temperatures at Amundsen–Scott station arise as a 0.01) (Fig. 4a) is consistent with the westerly wind increases over result of intrusions of warm air from the coastal region when recent decades. The largest increase in warm days was from the the planetary waves are amplified over the continent (see start of the record until the late 1990s, with a subsequent decrease. Sinclair 1981). The mean 500-hPa geopotential height anom- The percentage of days with cold conditions has not changed sig- alies for the days with the highest 5% of temperatures at the nificantly since the start of the record. station (Fig. 2f) has a positive anomaly close to the South Pole b. The eastern Antarctic Peninsula at 308E. Warm air can arrive at the South Pole from any sector of the continent, but as suggested in Fig. 2f, most incursions of Marambio has experienced a statistically significant increase 2 warm air arrive from close to the Greenwich meridian at a time (0.56% decade 1, p , 0.05) in the percentage of extreme warm when there is low pressure over the Weddell Sea (Clem et al. days over the record, which starts in the early 1970s. However, 2020). However, the highest temperature observed at the since the start of the twenty-first century the percentage of days

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2664 JOURNAL OF CLIMATE VOLUME 34

FIG. 4. Time series of the percentage of days each year with daily mean temperatures above the high temperature thresholds given in Table 3. with extreme warm temperatures has remained constant or de- The annual percentage of extreme warm days at Vernadsky is creased slightly, consistent with the summer surface temperatures, rather variable as illustrated by the switch in the number of ex- which increased markedly until the late 1990s, but have then de- tremes between 2018 (6.9%) and 2019 (0.6%) (Fig. 4c). As would creased since that time (Turner et al. 2016). At Marambio the be expected, most of the warm days in 2018 occurred during the number of cold days has decreased since the late 1970s at a rate summer months, but extreme high temperatures also occurred in 2 of 20.23% decade 1, although this trend was not significant. March, April, October, and November. The large number of warm days was a result of the greater number of deep storms to the west of c. The western Antarctic Peninsula the Antarctic Peninsula advecting warm air into the region. Bellingshausen and Vernadsky have both experienced an The five stations examined from the Antarctic Peninsula increase in the percentage of extreme warm days and decrease region all experienced a statistically significant (p , 0.01) in- in cold days over the full length of their temperature records. crease in extreme high temperatures in the late-twentieth- The decrease in the percentage of cold days at Vernadsky century part of their records, although the number of extremes 2 (21.67% decade 1, p , 0.05) (Fig. 5c) was the largest decrease decreased in subsequent years. in cold days of any of the stations and is consistent with the d. The eastern Weddell Sea large winter warming observed over the second half of the twentieth century and the loss of sea ice in midwinter (Turner Over the full length of the record Neumayer experienced a et al. 2013). There was also a significant increase (0.45% dec- decrease in the number of extreme high temperatures and in- 2 ade 1, p , 0.01) in the number of warm days, which was again crease in low temperatures, although neither trend was statis- most pronounced in the period up to 2000. tically significant.

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2665

FIG.5.AsinFig. 4, but showing the percentage of days each year with daily mean temperatures below the low temperature thresholds given in Table 3. e. East Antarctica warm days over the records extending back to the late 1950s. The positive trend in the percentage of warm days at Vostok Around the coast of East Antarctica there has been an overall small 2 (0.57% decade 1, p , 0.01) was the largest of any of the sta- reduction in the number of extreme warm and cold days at many of the tions outside of the Antarctic Peninsula. stations over the full length of the records (Table 4). The largest trends have been the decreases in the percentage of warm days at Casey 2 2 6. Conclusions (20.71% decade 1, p , 0.05), Mawson (20.37% decade 1, p , 2 0.10), and Molodezhnaya (20.91% decade 1, p , 0.05). These trends Our investigation of extreme temperatures in the Antarctic are consistent with the small decreases in summer temperatures at the has highlighted the importance of flow over orography in cre- stations, which has been linked to the loss of stratospheric ozone and ating the conditions that lead to record high temperatures. It the decrease in poleward heat flux. The largest decrease in cold days at has been well known for some time that the foehn effect plays the East Antarctic coastal stations has been at Novolazarevskaya an important part in establishing the conditions that give very 2 (20.42% decade 1, p , 0.05), which was dominated by the reduction high temperatures at some stations around the Antarctic over the 1960s to the 1990s. There has been no significant trend in the Peninsula. However, it has not previously been reported that SAM during the winter so the change must have occurred as a result of the highest temperatures at Mawson on the coast of East other factors. Antarctica occur when there is strong flow down the glacial valleys associated with a low to the east of a station and a ridge f. The Antarctic Plateau to the west. The air on the plateau inland of Mawson has a high The two stations on the Antarctic Plateau have both potential temperature and can also be warmed through descent experienced a significant positive trend in the percentage of of air in a midtropospheric ridge over the area. However, air

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2666 JOURNAL OF CLIMATE VOLUME 34

TABLE 4. Decadal trends in the percentage of extreme warm and cold days over the full length of the records and for 1979–2019. The significance of the trends is indicated with asterisks: *, p , 0.1; **, p , 0.05; ***, p , 0.01.

2 2 2 2 Warm (% decade 1), full Cold (% decade 1), full Warm (% decade 1), Cold (% decade 1), Station length length 1979–2019 1979–2019 Orcadas 0.60*** 20.05 0.19 20.29 Bellingshausen 0.22 21.05*** 20.55** 20.99 Esperanza 0.65*** 20.78*** 0.31 20.29 Marambio 0.54** 20.53** 0.28 20.23 Vernadsky 0.45*** 21.95*** 20.02 21.67** Casey 20.71** 20.19 20.68 20.14 Mirny 20.22 20.34* 20.32 20.33 Dumont d’Urville 0.07 20.16 20.42 20.15 Rothera 20.32 21.04 20.45 20.95 Mawson 20.37* 20.10 20.65* 20.47 Molodezhnaya 20.91** 20.21 21.25** 20.61 Davis 0.05 0.13 20.17 0.08 Syowa 20.06 20.26 20.31 20.05 Neumayer 20.16 0.23 20.16 0.23 Novolazarevskaya 0.12 20.42** 20.34 0.05 Vostok 0.57*** 0.06* 0.29 20.27 Amundsen–Scott 0.35* 0.02* 0.44 20.48 can also be forced up to the plateau on the western side of a since the start of the twenty-first century there has been little trend ridge in the coastal area before descent down the glacial val- in the SAM and local variability in the atmospheric circulation has leys. Our study has also highlighted the importance of sea ice in had a major effect on temperatures on the Antarctic Peninsula leading to record low temperatures. At Orcadas and the sta- (Turner et al. 2016). tions on the Antarctic Peninsula southerly flow leads to In situ observations, reanalysis datasets and ice-core climate northward advection of sea ice, which then limits the flux of records have shown that the atmospheric circulation around heat from the relatively warm ocean. the Antarctic Peninsula is highly variable on interannual and Over recent decades there has been an increase in the number decadal time scales (Connolley 1997). This makes it particu- of extreme high temperatures observed over Earth as a whole larly difficult to detect a signal of the impact of increasing (Alexander et al. 2006; Brown et al. 2008), with anthropogenic greenhouse gas concentrations. Whereas there is now strong influences being strongly linked to this upward trend (Fischer evidence that the global increase in extreme high temperatures and Knutti 2015). However, there are large regional variations in can be attributed to increasing greenhouse gases, local factors, these trends, with some areas having experienced a decrease in such as the loss of stratospheric ozone and changes in the ex- the number of warm days (Alexander et al. 2006). Clearly re- tent of sea ice, have been more important in driving the trends gional factors are playing a part and natural climate variability in extreme temperatures in the Antarctic over recent decades will mask some of the anthropogenic influences on these small (Polvani et al. 2011; Turner et al. 2013). scales. While a number of these global studies into extreme During the twenty-first century we anticipate a recovery of the temperatures have used data from the Antarctic stations to Antarctic ozone hole. However, if greenhouse gas concentrations consider change on a global basis, none has examined the trends continue to rise, we expect an increase in near-surface tempera- in extreme temperatures at specific Antarctic stations. tures across the Antarctic of several degrees by the end of this Overall, the trends in the frequency of warm and cold extreme century (Bracegirdle et al. 2008). With an overall shift of the temperatures across the Antarctic are similar to the trends seen in temperature distributions to the right we can therefore expect an the mean temperatures at the stations. The most pronounced sig- increase in the number of extreme high temperatures and de- nals being a tendency toward higher temperatures/more extreme crease in low temperatures over this period. This potential change warm days over the Antarctic Peninsula and a small cooling/more could be investigated using the output of global climate model extreme cold days around the coast of East Antarctica. This was experiments, where 6-hourly data would allow the construction of particularly the case from the start of the records until the late 1990s temperature distributions at various points over the twenty-first with a reduction in these trends during the twenty-first century. This century. The anticipated temperature changes will have the pattern of contrasting trends between the Antarctic Peninsula and greatest impact in the coastal areas of the continent and projec- the rest of the Antarctic is characteristic of the impact of the SAM tions of the frequency of extreme events will aid investigations on Antarctic temperatures (e.g., Marshall and Thompson 2016; into ice shelf collapse and changes in the terrestrial biota. Fogt and Marshall 2020) and has been seen in the ice core records as well as the station observations. During the second half of the Acknowledgments. This work forms part of the Polar Science twentiethcenturytherewasapositivetrendintheSAMduring for Planet Earth program of the British Antarctic Survey, summer (Marshall 2003), with the springtime loss of stratospheric Natural Environment Research Council. We are grateful to Dr. ozone contributing to this change from the early 1980s. However, Matthew Lazzara, Antarctic Meteorological Research Center,

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 1APRIL 2021 T U R N E R E T A L . 2667

Space Science and Engineering Center, University of Wisconsin– Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Madison, USA, for providing satellite imagery of the Mawson area. Configuration and performance of the data assimilation sys- tem. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/ Data availability statement. The individual Antarctic station 10.1002/qj.828. observations are available at ftp://ftp.bas.ac.uk/ssbsrc/SCAR_ Ding, Y. F., X. Cheng, X. C. Li, M. Shokr, J. W. Yuan, Q. H. Yang, and F. M. Hui, 2020: Specific relationship between the surface EGOMA/SURFACE/. The ECMWF reanalysis fields can be air temperature and the area of the Terra Nova Bay Polynya, found at https://www.ecmwf.int/en/forecasts/datasets/reanalysis- Antarctica. Adv. Atmos. Sci., 37, 532–544, https://doi.org/ datasets/era-interim. 10.1007/s00376-020-9146-2. REFERENCES Elvidge, A. D., I. A. Renfrew, J. C. King, A. Orr, and T. A. Lachlan-Cope, 2016: Foehn warming distributions in nonlin- Abram, N. J., R. Mulvaney, F. Vimeux, S. J. Phipps, J. Turner, and ear and linear flow regimes: A focus on the Antarctic M. H. England, 2014: Evolution of the Southern Annular Peninsula. Quart. J. Roy. Meteor. Soc., 142, 618–631, https:// Mode during the past millennium. Nat. Climate Change, 4, doi.org/10.1002/qj.2489. 564–569, https://doi.org/10.1038/nclimate2235. European Academies Science Advisory Council, 2013: Trends in Alexander, L. V., and J. Arblaster, 2009: Assessing trends in ob- extreme weather events in Europe: Implications for national served and modelled climate extremes over Australia in re- and European Union adaptation strategies. EASAC Policy lation to future projections. Int. J. Climatol., 29, 417–435 Rep. 22, 20 pp., https://easac.eu/fileadmin/PDF_s/reports_ https://doi.org/10.1002/JOC.1730. statements/Easac_Report_Extreme_Weather_Events.pdf. ——, and Coauthors, 2006: Global observed changes in daily cli- Fischer, E., and R. Knutti, 2015: Anthropogenic contribution to global mate extremes of temperature and precipitation. J. Geophys. occurrence of heavy-precipitation and high-temperature ex- Res., 111, D05109, https://doi.org/10.1029/2005JD006290. tremes. Nat. Climate Change, 5, 560–564, https://doi.org/10.1038/ Alvarez, J. A., and B. J. Lieske, 1960: The Little America blizzard of nclimate2617. May 1957. Proc. Symp. on Antarctic Meteorology, Melbourne, Fogt, R. L., and G. J. Marshall, 2020: The southern annular mode: VIC, Australia, Australian Bureau of Meteorology, 115–127. Variability, trends, and climate impacts across the Southern Barrand, N. E., D. G. Vaughan, N. Steiner, M. Tedesco, P. Kuipers Hemisphere. Wiley Interdiscip. Rev.: Climate Change, 11, Munneke, M. R. van den Broeke, and J. S. Hosking, 2013: https://doi.org/10.1002/wcc.652. Trends in Antarctic Peninsula surface melting conditions from Genthon, C., D. Six, V. Favier, M. A. Lazzara, and L. M. Keller, observations and regional climate modeling. J. Geophys. Res. 2011: Atmospheric temperature measurement biases on the Earth Surf., 118, 315–330, https://doi.org/10.1029/2012JF002559. Bozkurt, D., R. Rondanelli, J. C. Marin, and R. Garreaud, 2018: Foehn Antarctic Plateau. J. Atmos. Oceanic Technol., 28, 1598–1605, event triggered by an atmospheric river underlies record-setting https://doi.org/10.1175/JTECH-D-11-00095.1. temperature along continental Antarctica. J. Geophys. Res. IPCC, 2012: Managing the Risks of Extreme Events and Disasters to Atmos., 123, 3871–3892, https://doi.org/10.1002/2017JD027796. Advance Climate Change Adaptation. Cambridge University Bracegirdle, T. J., and G. J. Marshall, 2012: The reliability of Press, 582 pp. Antarctic tropospheric pressure and temperature in the latest Jones, D. A., and I. Simmonds, 1993: A climatology of Southern global reanalyses. J. Climate, 25, 7138–7146, https://doi.org/ Hemisphere extratropical cyclones. Climate Dyn., 9, 131–145, 10.1175/JCLI-D-11-00685.1. https://doi.org/10.1007/BF00209750. ——, W. M. Connolley, and J. Turner, 2008: Antarctic climate King, J. C., and J. Turner, 1997: Antarctic Meteorology and change over the twenty first century. J. Geophys. Res., 113, Climatology. Cambridge University Press, 424 pp. D03103, https://doi.org/10.1029/2007JD008933. ——, D. Bannister, J. S. Hosking, and S. R. Colwell, 2017: Causes of Brown, S. J., J. Caesar, and C. A. T. Ferro, 2008: Global changes in the Antarctic region record high temperature at Signy Island, extreme daily temperature since 1950. J. Geophys. Res., 113, 30th January 1982. Atmos. Sci. Lett., 18, 491–496, https:// D05115, https://doi.org/10.1029/2006JD008091. doi.org/10.1002/asl.793. Burton, B., 2014: Stevenson screen temperatures—An investiga- Kirchgaessner, A., J. King, and A. Gadian, 2019: The representa- ö tion. Weather, 69, 156–160, https://doi.org/10.1002/wea.2166. tion of f hn events to the east of the Antarctic Peninsula in Cape, M. R., M. Vernet, P. Skvarca, S. Marinsek, T. Scambos, and simulations by the Antarctic Mesoscale Prediction System. E. Domack, 2015: Foehn winds link climate-driven warming to J. Geophys. Res. Atmos., 124, 13 663–13 679, https://doi.org/ ice shelf evolution in Antarctica. J. Geophys. Res. Atmos., 120, 10.1029/2019JD030637. 11 037–11 057, https://doi.org/10.1002/2015JD023465. Marshall, G. J., 2003: Trends in the southern annular mode Cardil, A., D. M. Molina, and L. N. Kobziar, 2014: Extreme tem- from observations and reanalyses. J. Climate, 16, 4134– perature days and their potential impacts on southern Europe. 4143, https://doi.org/10.1175/1520-0442(2003)016,4134:TITSAM. Nat. Hazards Earth Syst. Sci., 14, 3005–3014, https://doi.org/ 2.0.CO;2. 10.5194/nhess-14-3005-2014. ——, and D. W. J. Thompson, 2016: The signatures of large-scale Clem, K., R. L. Fogt, J. Turner, B. R. Lintner, G. J. Marshall, J. R. patterns of atmospheric variability in Antarctic surface tem- Miller, and J. A. Renwick, 2020: Record warming at the South peratures. J. Geophys. Res. Atmos., 121, 3276–3289, https:// Pole during the past three decades. Nat. Climate Change, 10, doi.org/10.1002/2015JD024665. 762–770, https://doi.org/10.1038/s41558-020-0815-z. ——, A. Orr, N. P. M. van Lipzig, and J. C. King, 2006: The impact Comiso, J. C., and F. Nishio, 2008: Trends in the sea ice cover using of a changing Southern Hemisphere annular mode on Antarctic enhanced and compatible AMSR-E, SSM/I and SMMR data. Peninsula summer temperatures. J. Climate, 19, 5388–5404, J. Geophys. Res., 113, C02S07, https://doi.org/10.1029/2007JC004257. https://doi.org/10.1175/JCLI3844.1. Connolley, W. M., 1997: Variability in annual mean circulation in Nicolas, J. P., and D. H. Bromwich, 2011: Climate of West southern high latitudes. Climate Dyn., 13, 745–756, https:// Antarctica and influence of marine air intrusions. J. Climate, doi.org/10.1007/s003820050195. 24, 49–67, https://doi.org/10.1175/2010JCLI3522.1.

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC 2668 JOURNAL OF CLIMATE VOLUME 34

Parkinson, C. L., 2014: Global sea ice coverage from satellite data: warming on the western Antarctic Peninsula. Int. J. Climatol., Annual cycle and 35-yr trends. J. Climate, 27, 9377–9382, 33, 852–861, https://doi.org/10.1002/joc.3474. https://doi.org/10.1175/JCLI-D-14-00605.1. ——, and Coauthors, 2016: Absence of 21st century warming on Polvani, L. M., D. W. Waugh, G. J. P. Correa, and S.-W. Son, 2011: Antarctic Peninsula consistent with natural variability. Nature, Stratospheric ozone depletion: The main driver of twentieth-century 535, 411–415, https://doi.org/10.1038/nature18645. atmospheric circulation changes in the Southern Hemisphere. ——, T. Phillips, G. J. Marshall, J. S. Hosking, J. O. Pope, T. J. J. Climate, 24, 795–812, https://doi.org/10.1175/2010JCLI3772.1. Bracegirdle, and P. Deb, 2017: Unprecedented springtime Scambos, T. A., J. A. Bohlander, C. A. Shuman, and P. Skvarca, retreat of Antarctic sea ice in 2016. Geophys. Res. Lett., 44, 2004: Glacier acceleration and thinning after ice shelf collapse 6868–6875, https://doi.org/10.1002/2017GL073656. in the Larsen B embayment, Antarctica. Geophys. Res. Lett., ——, and Coauthors, 2019a: The dominant role of extreme precip- 31, L18402, https://doi.org/10.1029/2004GL020670. itation events in Antarctic snowfall variability. Geophys. Res. ——, G. G. Campbell, A. Pope, T. Haran, A. Muto, M. Lazzara, Lett., 46, 3502–3511, https://doi.org/10.1029/2018GL081517. C. H. Reijmer, and M. R. van den Broeke, 2018: Ultralow ——, G. Marshall, K. R. Clem, S. R. Colwell, T. Phillips, and H. Lu, surface temperatures in East Antarctica from satellite thermal 2019b: Antarctic temperature variability and change from infrared mapping: The coldest places on Earth. Geophys. Res. station data. Int. J. Climatol., 40, 2986–3007, https://doi.org/ Lett., 45, 6124–6133, https://doi.org/10.1029/2018GL078133. 10.1002/joc.6378. Schlosser, E., B. Stenni, M. Valt, A. Cagnati, J. G. Powers, K. W. van den Broeke, M. R., 1998: The semi-annual oscillation and Manning, M. Raphael, and M. G. Duda, 2016: Precipitation and Antarctic climate. Part 1: Influence on near surface tempera- synoptic regime in two extreme years 2009 and 2010 at Dome C, tures (1957–79). Antarct. Sci., 10, 175–183, https://doi.org/ Antarctica—Implications for ice core interpretation. Atmos. Chem. 10.1017/S0954102098000248. Phys., 16, 4757–4770, https://doi.org/10.5194/acp-16-4757-2016. Vavrus, S. J., J. E. Walsh, and W. L. Chapman, 2006: The behavior Sinclair, M. R., 1981: Record-high temperatures in the Antarctic—A of extreme cold air outbreaks under greenhouse warming. Int. synoptic case study. Mon. Wea. Rev., 109, 2234–2242, https:// J. Climatol., 26, 1133–1147, https://doi.org/10.1002/joc.1301. doi.org/10.1175/1520-0493(1981)109,2234:RHTITA.2.0.CO;2. Wake, L. M., and S. J. Marshall, 2015: Assessment of current Skansi, M., and Coauthors, 2017: Evaluating highest temperature methods of positive degree-day calculation using in situ ob- extremes in the Antarctic. Eos, Trans. Amer. Geophys. Union, servations from glaciated regions. J. Glaciol., 61, 329–344, 98, https://doi.org/10.1029/2017EO068325. https://doi.org/10.3189/2015JoG14J116. Thompson, D. C., D. W. Limbert, and G. Manley, 1970: The coreless Wei, T., Q. Yan, and M. Ding, 2019: Distribution and temporal winter at Scott Base, Antarctica. Quart. J. Roy. Meteor. Soc., 96, trends of temperature extremes over Antarctica. Environ. Res. 556–557, https://doi.org/10.1002/qj.49709640921. Lett., 14, 084040, https://doi.org/10.1088/1748-9326/ab33c1. Turner, J., and Coauthors, 2004: The SCAR READER project: Zazulie, N., M. Rusticucci, and S. Solomon, 2010: Changes in cli- Towards a high-quality database of mean Antarctic mete- mate at high southern latitudes: A unique daily record at orological observations. J. Climate, 17, 2890–2898, https:// Orcadas spanning 1903–2008. J. Climate, 23, 189–196, https:// doi.org/10.1175/1520-0442(2004)017,2890:TSRPTA.2.0. doi.org/10.1175/2009JCLI3074.1. CO;2. Zhai, P., and X. Pan, 2003: Change in extreme temperature and ——, and Coauthors, 2009: Record low surface air temperature at precipitation over northern China during the second half of Vostok station, Antarctica. J. Geophys. Res., 114, D24102, the 20th century. Acta Geogr. Sin., 58, 1–10. https://doi.org/10.1029/2009JD012104. Zwally,H.J.,J.C.Comiso,C.L.Parkinson,D.J.Cavalieri,and ——, T. Maksym, T. Phillips, G. J. Marshall, and M. P. Meredith, P. Gloersen, 2002: Variability of Antarctic sea ice 1979–1998. 2013: Impact of changes in sea ice advance on the large winter J. Geophys. Res., 107, 3041, https://doi.org/10.1029/2000JC000733.

Unauthenticated | Downloaded 09/30/21 03:34 PM UTC